* Author to whom correspondence should be addressed. Computers Chem. Engng Vol. 22, No. 11, pp. 1673—1685, 1998  1998 Published by Elsevier Science Ltd All rights reserved. Printed in Great Britain PII: S0098-1354(98)00228-2 0098—1354/98 $ — see front matter Automating operating procedure synthesis for batch processes: Part I. Knowledge representation and planning framework Shankar Viswanathan, Charlotta Johnsson, Rajagopalan Srinivasan, Venkat Venkatasubramanian* and Karl Erik A rzen  Laboratory for Intelligent Process Systems, School of Chemical Engineering, Purdue University, West Lafayette, IN 47907, USA  Department of Automatic Control, Lund Institute of Technology, Box 118, S—221 00 LUND, Sweden (Received 14 April 1997; revised 6 February 1998) Abstract Automating the synthesis of operating procedures for batch processes is very valuable as plant personnel often spend considerable amount of time and effort in preparing and verifying them for correctness and completeness. Towards this goal, a framework for automating operating procedure synthesis for batch processes is proposed in this paper. We propose an approach based on Grafcet, a discrete event modeling concept, to represent procedural knowledge combined with an object-oriented representation of the declarative knowledge. Grafcet is also used to model the information; called inferred knowledge, that is incrementally generated during operating procedure synthesis. The details of the basic components and the advanced features of Grafcet that make it a suitable technique for modeling the procedural and inferred knowledge are presented. A hierarchical planning strategy is proposed that uses the declarative and procedural knowledge to generate the inferred knowledge incrementally, which leads to the synthesis of the operating procedures. The implementa- tion of this framework and its application to an industrial case study are presented in Part II.  1998 Published by Elsevier Science Ltd. All rights reserved. 1. Introduction Production of chemicals in batches is a dominant and common mode of operation for production of low-volume, high value products such as pharmaceut- ical chemicals. To successfully manage batch produc- tion many control functions need to be implemented. The control activity model shown in Fig. 1 identifies the major batch control activities and the relation- ships amongst them. This model, outlined in the batch S88 standards (Instruments Society of America, 1995), provides an overall perspective of batch process man- agement. The problem of synthesizing operating pro- cedures is a combination of two of these batch control activities, namely, recipe management, and production planning and scheduling. Operating procedures are the detailed sequence of instructions an operator needs to manage a batch process safely and optimally. Operating procedures can be generated off-line before production begins or online as production information becomes available. In this paper, we address the problem of off-line oper- ating procedure synthesis, which is the commonly used approach. In off-line operating procedure synthesis, which is done prior to actual production, the information available can be classified into information about the process and about the plant in which the production is to be done. The information about the process is called the recipe. As the operating procedures are generated, the recipe goes through modifications and is variously referred to, in the S88 standards, depend- ing on its information content, as the general, site, master, and control recipe. Initially the recipe, called general recipe, contains information about how a chemist produced the desired product on a laborat- ory scale. It gives details about the materials involved in, and the operating conditions of, a sequence of reactions and unit operations. By incorporating site- specific variances, such as local raw material differ- ences, the information in the general recipe is modified 1673